How to Identify In-Market B2B Buyers Using Intent Data and Buying Signals

Identify in-market B2B buyers by scoring accounts against four kinds of buying signal firmographic fit, third-party intent, first-party engagement, and relationship change then weighting them by recency and velocity instead of raw count. In a joint 2026 study of 243,000 G2 Buyer Intent accounts across 11 cybersecurity programmes, accounts appearing in both intent data and a demand programme generated 21% more leads per account, 93% higher multi-touch engagement, and were 1.6 times more likely to be at decision stage.

The vendor matters less than the scoring rules on top of it. A taxonomy, a decay function and a tiering scheme with an SLA beat another feed of contacts.

93% more multi-touch engagement for accounts seen in both G2 Buyer Intent and a demand programme INFUSE/G2 study (B2BMX Insights, July 2026).

Why in-market detection got harder in 2026

Most of a category’s buyers are not shopping right now. The B2B Institute’s 95-5 rule, published through LinkedIn Marketing Solutions, puts roughly 95% of potential buyers “out-market” at any moment, and its joint research with the Ehrenberg-Bass Institute shows how narrow that window is.

The 5% who are in market have moved their research where you cannot see it. G2’s March 2026 survey of 1,076 B2B software buyers found 51% now begin research in an AI chatbot more often than with Google, up from 29% a year earlier, and 71% rely on AI chatbots somewhere in the process. TrustRadius’s 2026 B2B Buying Disconnect Report (1,862 buyers, 444 vendors, surveyed January 2026) shows where that research lands: buyers ranked product demos, free trials and user reviews as most influential, and analyst reports were used by just 13% of buyers, a 63% drop since 2022.

Most of the decision also happens outside your sight. Consensus’s 2026 B2B Buyer Behavior Report, built on more than 6 million anonymised customer interactions, found buyers spend only about 17% of the purchase journey in direct contact with potential suppliers, and Ipsos’s 2026 B2B Buyer Journey Report (288 US decision-makers, August 2026) put 77% saying vendor research is the most time-consuming stage of buying.

So “is this account in market?” needs multiple kinds of evidence, fast. This is the gap between demand generation and lead generation that most teams never close: one hands you a list of contacts, the other tells you which accounts deserve a rep’s time.

Identify in-market B2B buyers with a four-layer signal taxonomy

Intent data is the record of a buyer’s observable research behaviour what they searched for, read, compared, and came back to. A workable in-market process splits that record into four layers, because each answers a different question and carries a different error rate.

INFUSE’s Voice of the Buyer 2026 splits the problem three ways eligibility from firmographic data, timing from intent, readiness from buying group activity. I use four, because relationship change is the layer most teams ignore. Madison Logic’s 2026 B2B Marketing Outlook makes the same point: intent now arrives as research behaviour, engagement patterns and account-level signals at once.

Layer 1: Firmographic fit signals can they buy, not are they buying

Company size, industry, revenue band, geography, and confirmed technology fit. Treat this layer as a filter rather than a score: it answers whether the account could buy at all, and it should hard-stop scoring when the answer is no. The trap is treating fit as interest: a dormant account and a ready buyer with the same profile look identical on firmographics alone.

Accuracy is improving fast here, and it is the layer where vendors are most auditable. In its August 2026 product release notes, 6sense said it had surpassed 90% accuracy on core firmographic fields company size, revenue, industry, location for public and visible companies. That is vendor-stated, so ask for the measurement.

Layer 2: Behavioral intent signals they are researching something

Third-party intent is where most teams begin, and it is the layer with the most noise: mechanics differ enough between vendors that accuracy differs too.

Bombora describes a co-operative of hundreds of publishers and thousands of B2B media sites, where 86% of the data is shared exclusively with Bombora for deriving intent, and where its tag captures engagement inside gated, paywalled and open content across more than 25,300 tracked topics. Demandbase’s documentation describes the same problem from the other side: it identifies buying signals from keywords in web searches, maps them to decision-makers, and “anonymizes browsing activity while retaining the digital footprint by which you can track and reach these stakeholders.”

What Layer 2 proves: a category or problem is on a screen somewhere now. What it does not prove: who that person is, or that they are shopping from you.

Layer 3: Engagement signals they engaged with you

First-party behaviour on your own property: repeat visits, pricing and comparison pages, gated asset downloads, demo requests, webinar attendance, email engagement, chat, time in product documentation. This layer carries the most weight because the behaviour is about you specifically, and your own server logs know more about a session than any third party does.

Layer 4: Relationship signals what changed, and who decides

CRM stage and history, champion tenure, a new hire landing in a buying role, a technology-stack change, a funding round, a partner referral. These explain why now, and they route the account to the right human.

The layer most teams skip is role coverage. Forrester’s 2026 State of Business Buying found 13 internal stakeholders and nine external participants influence buying decisions on average, and buying groups for purchases with genAI features run double the size of those without 14 members against seven. INFUSE’s Voice of the Buyer 2026 puts the global average buying group at nine, rising to ten in EMEA.

First-party intent and third-party intent do different jobs

Mixing the two up is a common cause of a bought intent programme that produces noise instead of pipeline. One finds the research; the other confirms the engagement.

Intent-based targeting that ignores the split produces noise.

The rule I would defend to a CFO: no account reaches your top tier on third-party intent alone. Layer 2 opens the door; Layer 3 confirms you were in the running. Our guide to the best B2B intent data providers covers the vendor landscape.

Turning anonymous website visitors into accounts

Website visitor identification resolves an anonymous web session to a company or a person. It works in two tiers that are not equally trustworthy.

Deterministic identification is the easy half. A form fill, a signup, a login or a calendar booking gives you a verified person and, through enrichment, an account.

Probabilistic identification is the hard half. Mapping an IP address to a company, inferring a firm from geography and device fingerprint, or tying a session to an account through a shared cookie or identity graph. These produce a hypothesis, not a fact.

DemandScience’s September 2026 assessment is blunt about why: bidstream-aggregated data cannot verify identity or the authenticity of intent, so competitors, interns and anonymous visitors all surface as false positives.

Two rules follow. Publish your match rate internally, because a tool resolving 4% of traffic and one resolving 18% produce very different tiering math. And carry every probabilistic match’s evidence pages, timestamps, inferred company into the CRM, so a rep can judge it rather than trust it.

Account intent scoring: velocity and recency beat a raw score

A score is a snapshot. What matters is the change in the score, because a buyer moving from nothing to something in a week is different from an account parked at a middling number for six months. The INFUSE/G2 study supports that: across 11 programmes, the average overlap rate between G2 Buyer Intent accounts and demand-programme-engaged accounts was 42%, ranging from 22% to 90%, and 57.4% of overlap accounts reached three or more content touches against 29.7% of non-overlap accounts.

The point values below are a starting model, not a research finding. Calibrate them against your own closed-won data.

Five rules turn those points into a score:

• Recency decay. Points fall linearly to zero at twice the window.
• Velocity multiplier. Multiply by 1.25 when three or more signal events land inside seven days.
• Breadth multiplier. Multiply by 1.25 when two or more roles engage within 14 days.
• Fit gate. If Layer 1 fails, the score is zero regardless of intent.
• Single-contact cap. If the score rests on one known individual, cap the account at Tier 2 until a second role appears.

Tier accounts by signal depth, not by score alone

Account tiering decides who gets a named rep, who gets a nurture sequence, and who gets nothing. Tier on signal depth how many signal types, across how many layers, involving how many people not on the score alone.

Keep Tier 1 small enough that the SLA is real: a tier promising four hours that averages two days is worse than one promising two days, because reps stop believing the system. And remember a Tier 1 account is a buying group, not a contact. The INFUSE/G2 study reports buying cycles compressing to seven months, and Consensus’s 2026 research found demo content shared with seven or more stakeholders tied to an opportunity 36% of the time against 19% for content shared with nobody.

Wider account-selection logic belongs in your account-based marketing strategy; the tier boundary belongs to whoever owns the SLA.

How to build the in-market process: six steps

1. Fix the eligibility gate first. Write down your ICP and technographic rules, and what happens when an account fails them.
2. Cap the topic taxonomy. Choose a few topic areas describing problems you actually solve; broad taxonomies inflate volume and destroy precision.
3. Map every data source to a layer. If you cannot say which layer a source feeds, it is a contact list, not a signal.
4. Write the decay rules before the points. If the window is decided after you see which accounts scored well, you fitted the model to the outcome.
5. Build identity resolution and publish its hit rate. Track monthly what share of traffic resolves to an account, and to a person. If you cannot state the number, your scoring is guessing.
6. Set tier and SLA together, then hand over evidence. Show the page, the date, the source never a score and review false positives monthly.

Handing signals to sales without destroying trust

Reps forgive a signal that misses. They do not forgive one that lies.

The strongest evidence comes from a practitioner rather than a vendor. Tara Faulkner of Informa TechTarget, quoted in a 2026 B2BMX Insights feature: “Don’t spend another dime on intent without taking a hard look at what happens to the signal you already have. A strong signal fumbled by poor processes produces the same result as a bad one.” Her numbers make the point: at her last company, 80% of prospects who took a meeting became real opportunities, while only 13% of flagged leads ever reached that meeting.

Three rules make that work in practice:

• Send the evidence, not the score. “Viewed our pricing page three times between the 12th and 16th” is actionable. “Score 82” is a task a rep cannot verify.
• Keep one owner and one clock. Every tier needs a named owner and a published response window, or the signal queue becomes a graveyard. Wire it into the routing rules in marketing operations routing; the lead routing SOP covers the SLA.
• Never let a signal override the CRM. An open opportunity, a closed-lost reason, a customer in renewal: the CRM outranks the score, and your model should be barred from re-engaging accounts a rep has disqualified.

Then match the outreach to the buying stage. Buying stage signals a pricing page, a booked meeting, a competitor comparison should decide the message, and Tier 1 accounts at decision stage want validation material, not another explainer. That is the argument for GTM messaging architecture.

Demandbase’s October 2026 research found 72% of B2B marketing teams the same size or smaller than two years ago while 70% report shorter timelines, which is why the list must shrink before it reaches a rep.

Where intent data goes wrong

Intent data produces false positives, regardless of the vendor. DemandScience’s September 2026 assessment sorts intent data into three tiers: bidstream-aggregated data, cheapest and broadest but unable to distinguish research from incidental traffic; behavioural research data from research platforms, more reliable but vendor-dependent; and peer-review behavioural data, where identity is verified, with the lowest false-positive rate and the smallest coverage.

Two traps matter. Across the 11 programmes in the INFUSE/G2 study, roughly 196,000 G2 Buyer Intent accounts were never targeted by an active demand programme, so a signal you never act on looks identical to one that does not exist. And with 13 internal stakeholders plus nine external participants influencing a typical decision, single-contact scoring is structurally incomplete.

None of this argues against intent data. It argues for scoring it, tiering it, and shrinking the list until it is short enough to work.

Frequently asked questions

How do you identify in-market B2B buyers?

Score the account against four signal layers, then apply recency decay and a velocity multiplier before tiering by signal depth. The strongest published evidence is the joint INFUSE and G2 study from July 2026: 21% more leads per account, 93% more multi-touch engagement, and 1.6 times the likelihood of reaching decision stage.

What is intent data in B2B marketing?

Intent data is the record of a buyer’s observable research behaviour what they looked at, searched for, and returned to. It comes in two forms: third-party intent, gathered on publisher networks and review sites, and first-party intent, gathered on your own site and email.

What are the strongest buying signals?

Separated by role: comparison and competitor research, gated asset downloads tied to a named person, booked meetings, and engagement by two or more distinct roles. In the INFUSE/G2 study, 80% to 98% of overlap accounts showed competitive signal behaviour, and accounts recording competitor pageviews generated up to 41% more leads per company. Two roles engaging inside 14 days outranks any single action only it speaks to the buying group.

How do you identify anonymous website visitors and turn them into accounts?

Start with deterministic signals form fills, logins, calendar bookings because they give you a verified person and, through enrichment, an account. Add probabilistic resolution as a second pass, using IP-to-company mapping, device and geography fingerprints, or identity-graph matching, and treat every result as a hypothesis. DemandScience’s 2026 assessment is blunt that bidstream data cannot verify identity or the authenticity of intent, and Tara Faulkner of Informa TechTarget frames the test as two questions: can you name the person, and did the engagement happen?

How many signals do you need before you call an account in-market?

One strong signal gets you into Tier 2; two signal types across two layers with two or more roles engaged gets you into Tier 1. No published research sets a universal threshold the INFUSE/G2 study reports an overlap rate of 42% on average but ranging from 22% to 90% across programmes. Calibrate against your own data.

What is the difference between intent data and website traffic data?

Traffic data tells you who came to your site. Intent data tells you what was being shopped. They overlap only where a real person did both, and that overlap is the highest-quality signal available: an account that visited your pricing page three times in five days beats any topic score.

How do you combine first-party and third-party intent data?

Use third-party intent for discovery and timing, and first-party intent for confirmation and intent strength. Layer 2 opens the account and Layer 3 decides the tier. Track them as separate inputs so you can see which one carried the decision.

What to do next

• Write down your Layer 1 eligibility rules before you evaluate any intent vendor.
• Audit your last 50 flagged accounts by hand that is your real accuracy.
• Pick the SLA for Tier 1 and put a name against it.